> For the complete documentation index, see [llms.txt](https://dots.gitbook.io/dots-docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://dots.gitbook.io/dots-docs/analysis/highlights-explorer.md). # Highlights Explorer The Highlights Explorer is the dashboard where all of your highlights come together in one place. Once highlights have been created across your documents, whether manually, with AI assistance, or through auto highlights, this is where you step back and explore them across your entire dataset rather than one document at a time. The Highlights Explorer gathers every highlight into a single, searchable, filterable view. In the platform you will find it under ***Highlights*** in the main left panel. #### **Why use the Highlights Explorer?** A single study can produce thousands of highlights spread across hundreds of documents. Reading them one by one can become overwhelming; but gathered together and filtered the right way, they become easier to explore and analyze. The Highlights Explorer allows you to : * **View highlights** from across your datasets together in one place * **Filter** highlights by tags, themes, attributes, source, date, author, sentiment, or any other available fields. * **Examine** how common a theme is, which themes occur together, and how sentiment maps onto topics. * Generate **summaries** of your preferred filtered view or export them for further analysis. Using our WASH example: instead of rereading every India and Kenya interview, you can examine the prevalence count of *Shared Community Access* across countries, whether it co-occurs with negative sentiment often, and overall what patterns emerge across both countries. ### Understanding the interface
The Explorer has a three-part layout: * A **left sidebar** for Filters and Color Code. * A **main area in the center**, with a search and sort bar at the top, highlight cards below, and controls for count, export, and AI summary just above them. * A **right-hand visualization panel** for deeper analysis using Prevalence and Co-occurrence. #### Highlight Cards {% columns %} {% column width="66.66666666666666%" %} Each highlight appears as a card showing * the highlighted **text segment** * the **source document** it came from * its **tags** (mentioned more below) * a badge indicating the **content type** (interview, survey response, support ticket, and so on). Tags on the card are split into: * annotation tags applied directly to the highlight and * document tags it inherits from its source document. Click any card to open the highlight in its source document and jump straight to the relevant passage. {% endcolumn %} {% column width="33.33333333333334%" %}
{% endcolumn %} {% endcolumns %} {% hint style="info" %} *The grid loads additional highlights as you scroll.* {% endhint %} #### Filtering and searching Filters are the engine of the explorer. Whatever you filter to, everything else respects that selection: counts, charts, summaries, and exports all reflect the slice of highlights you are currently viewing. Filters are generated automatically from the content templates in your workspace, so the dimensions available reflect your own data. {% columns %} {% column width="66.66666666666666%" %} A common workflow is: * Filter to the subset of highlights you want to explore * Review the matching highlights * Analyze patterns * Generate summaries or export the results Typically you can filter by: * tags and themes * attributes * any fields from the source documents * content type of source document * date ranges * annotation & document author You can select multiple values within a filter and combine them with AND / OR logic to build very specific slices. For example, *"I want to search for highlights of Sanitation Challenges + only about Behavior Change + highlights without the positive sentiment"* (referring to the image on the side). {% endcolumn %} {% column width="33.33333333333334%" %}
{% endcolumn %} {% endcolumns %} At the top of the explorer, a keyword search allows you to look across the highlights, while the sort control allows you to order results by Latest, Oldest, or A–Z. #### Analyzing Highlights: Prevalence & Co-occurrence Filtering and browsing help you focus on the highlights you want to understand. The explorer also provides analysis tools to help you dive deeper into the patterns you find. **Prevalence: how common is each theme?** {% columns %} {% column width="66.66666666666666%" %} Prevalence answers the question \ *“How often does each tag or theme appear in what I’m looking at?”* It is shown as bar charts grouped by theme, with a toggle that allows you to switch between thematic and attribute tags. A tag’s percentage is calculated by the number of highlights carrying it divided by the total highlights currently in scope. Clicking any bar drills directly into the highlights behind it. For example, across your filtered set you might see that *Shared Community Access* appears in 40% of highlights while *Handwashing* appears in 15%. That gives you an immediate sense of which topics dominate the conversation. {% endcolumn %} {% column width="33.33333333333334%" %}
{% endcolumn %} {% endcolumns %} **Co-occurrence: which themes travel together?**
Co-occurrence answers the question \ *“Which tags appear together, and is that meaningful or just noise?”* It is displayed as a table of tag pairs, each with four measures:
ReadingWhat it tells you
CountHow many highlights contain both tags? This suggests the raw volume behind the pairing.
StrengthDo these tags appear together more often than you would expect by chance? Higher values suggest a stronger relationship between the tags. For example, 1 and below means coincidence; above 1 (roughly 1.2+) means a real association.
OverlapHow much do the highlights for these two tags overlap? Very high overlap often means the two tags are near-synonyms rather than a meaningful pairing.
PredictabilityIf one tag appears, how often does the other appear as well? Read it as: when A appears, B appears X% of the time.
By default, co-occurrence pairs thematic tags with attribute tags (for example, topics against sentiment). However, you can change the scope to thematic-only, attribute-only, or all tags. Click any row to automatically apply both tags as filters. To make analysis easier, the Explorer includes preset lenses: ready-made views with sensible noise filters already built in, such as: * ***Top topic pairings*** * ***Negative topic pulse*** and ***Positive topic pulse*** (topics most tied to negative or positive sentiment) * Several ***signal-focused views***. These presets are a good starting point before you build your own. Additionally there are toggle buttons available: * ***Show Columns*** - Toggles the additional co-occurrence columns (Strength, Overlap, and Predictability) on or off. Turn it off if you only want to see the Count column for a simpler view, and on when you want the full picture behind each pair. * ***Remove Filters*** - Applies the filters set in the left-hand filter panel to all of the visualizations, so the charts and tables reflect the same slice you've filtered the highlights down to. In our WASH example, co-occurrence might show that *Shared Community Access* and *Barrier* appear together far more often than chance would predict. That is strong evidence that shared facilities are experienced as a real problem, rather than simply being mentioned in passing. **Split: prevalence and co-occurrence together** The Split view shows Prevalence and Co-occurrence in the same visualization panel. It contains all the features of both views, combined into a single tab, so you can analyse how common a theme is and what it co-occurs with without moving between tabs. {% hint style="info" %} *The right-hand visualization panel containing Prevalence, Co-occurrence & Split is an advanced feature that may need to be enabled for your workspace. If you do not see it, contact your Dots representative.* {% endhint %} #### Additional features that help day-to-day **AI Summaries**
The Summarize button generates a narrative summary (roughly 250–500 words) of the highlights matching your current filters. It works only from the highlights actually in scope and does not invent quotes, numbers, regions, or personas, giving you a grounded overview of what your data is saying. The summary can be refreshed to regenerate, and results are cached so repeat views load quickly. **CSV Export**
The Export button downloads every highlight that matches your search criteria and current filters as a CSV file, ready for spreadsheet-based analysis, reporting or sharing. **Color Code Breakdown**
Select a tag category and assign a distinct color to each of its tags, so the highlights are visually segmented at a glance. #### Tips for getting the most out of Highlights Explorer * Start broad, then narrow. Look at everything first before using filters to zoom into the specific slices you want to understand. * Use prevalence to identify what matters, and co-occurrence to understand why. Prevalence tells you what is common while co-occurrence helps you understand what is connected. * Let the presets do the heavy lifting. They already filter out much of the noise, making them a reliable place to start . * Read strength and overlap together. A high-strength pair with very high overlap may simply represent two names for the same idea ; a high-strength pair with moderate overlap is usually the more interesting finding. * Sense-check the AI summary against the underlying highlights before quoting it. Although it is grounded in your data, treat it as a starting narrative rather than a final conclusion. * Export filtered results when you need deeper analysis, your own counts, custom pivots, or sharing outside the platform.